Related Experiment Videos
Random-effect models for ordinal responses: application to self-reported disability among older persons.
1INSERM Unité 500, 39, avenue Charles-Flahault, 34093 Montpellier. carriere@montp.inserm.fr
Revue D'Epidemiologie Et De Sante Publique
|April 13, 2006
Summary
This study introduces random-effects models for analyzing longitudinal data with ordinal outcomes, improving understanding of how risk factors like pain affect disability progression over time.
Area of Science:
- Epidemiology
- Biostatistics
- Gerontology
Background:
- Longitudinal studies with repeated ordinal outcomes are common in research.
- Analyzing such data presents challenges in model selection and handling within-subject correlations.
Purpose of the Study:
- To propose a fitting strategy for random-effects models in longitudinal studies with ordinal outcomes.
- To detail various ordinal models extended for repeated responses and their impact on random effect structures.
Main Methods:
- Utilized random-effect models for longitudinal data analysis.
- Applied model selection criteria and goodness-of-fit measures.
- Demonstrated with a seven-year study of self-reported disability in older women.
Main Results:
- Validated proportional odds for age and gait speed.
- Found that pain's impact varied by disability level.
- Identified the restricted partial proportional odds model as having good fit.
Conclusions:
- Random-effects models effectively account for ordinal outcomes in longitudinal studies.
- Allows for modeling risk factor impact across different response levels.
- Offers a valuable approach for understanding complex evolutionary processes.